Industrial Control Configuration Using Deep Learning Under Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing industrial control systems struggle to improve product quality beyond historical best without violating constraints, especially under changing environments with noisy data, relying on random experiments or domain experts for process parameter definition.
Innovation Solution
A deep learning-based method using a multilayer perceptron model to learn relationships among operating parameters, quality indicators, and constraints, employing gradient descent and soft ball optimization to predict optimal controllable parameter values, ensuring gradual improvements and adherence to constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If existing techniques like Bayesian optimization or gradient-based methods are used, then initial settings or historical data can be utilized, but they cannot consistently improve product quality beyond historical best under changing environments with noisy data
Solution Approach 1:
The patent segments the optimization process into distinct phases: data preprocessing phase, MLP model training phase, and gradient descent optimization phase. This segmentation allows each phase to handle specific challenges (noise filtering, relationship learning, precise optimization) independently, improving overall reliability of quality improvement
Solution Approach 2:
The patent introduces an MLP model as an intermediary between historical data and optimization. The MLP learns the underlying relationships in noisy data and provides a smoothed, reliable objective function for gradient descent, enabling consistent quality improvement beyond historical best
2Adaptability or versatility
If random experiments or domain expert methods are used to define process parameters, then initial configurations can be obtained, but the system cannot adapt to changing operating conditions and raw material quality
Solution Approach 1:
The system performs self-service by automatically learning relationships from historical data through the MLP model and autonomously optimizing parameters using gradient descent. This eliminates the need for domain experts to manually define parameters while adapting to changing conditions, achieving high adaptability without proportionally increasing operational complexity
Solution Approach 2:
The patent dynamically changes optimization parameters (learning rate, batch size, regularization strength) during the gradient descent process based on performance feedback. This allows the system to adapt to changing operating conditions automatically, maintaining versatility while managing complexity through automated parameter tuning
3Manufacturing precision
If gradient-based techniques are used to improve product quality beyond historical data, then optimization can be performed, but they do not perform efficiently when data is noisy
Solution Approach 1:
The patent performs preliminary data preprocessing and MLP model training before applying gradient descent optimization. This preliminary action filters out noise and learns robust relationships from historical data, enabling gradient-based techniques to work efficiently even when raw data is noisy, thereby improving both quality and measurement precision
Solution Approach 2:
The system implements feedback through the MLP model that continuously monitors the relationship between operating parameters and quality metrics. This feedback mechanism allows the gradient descent to adjust its path based on learned patterns from noisy data, maintaining efficient optimization while being robust to measurement imprecision
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
The embodiments of present disclosure herein address unresolved problem of getting optimum quality of product while changing operating conditions and raw material frequently in an industrial manufacturing process. The disclosure herein generally relates to a deep learning based approach for a multi-objective constrained optimization. Embodiments provide a method and system for recommending optimum configurations in industrial control systems for improving quality of product. The system is configured to automate improvement over existing golden batch with a data driven approach and replace need of random experimentation with very minimal systematic experiments. The system ensures that no constraint violations are made, and the system remains stable even when data is noisy. Further, the system recommends values of parameters so that improvement in quality is achieved. The changes in parameter value should be gradual even if the historical data received from feedback is noisy.